CN113447905B - Double millimeter wave radar human body fall detection device and detection method - Google Patents
Double millimeter wave radar human body fall detection device and detection method Download PDFInfo
- Publication number
- CN113447905B CN113447905B CN202110728497.9A CN202110728497A CN113447905B CN 113447905 B CN113447905 B CN 113447905B CN 202110728497 A CN202110728497 A CN 202110728497A CN 113447905 B CN113447905 B CN 113447905B
- Authority
- CN
- China
- Prior art keywords
- human body
- millimeter wave
- time
- layer
- frequency
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 57
- 238000000605 extraction Methods 0.000 claims abstract description 23
- 238000012545 processing Methods 0.000 claims abstract description 18
- 230000009466 transformation Effects 0.000 claims description 24
- 238000000034 method Methods 0.000 claims description 23
- 230000004927 fusion Effects 0.000 claims description 12
- 238000011176 pooling Methods 0.000 claims description 11
- 238000005070 sampling Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000015572 biosynthetic process Effects 0.000 claims description 2
- 230000010355 oscillation Effects 0.000 claims description 2
- 230000009977 dual effect Effects 0.000 claims 1
- 230000009471 action Effects 0.000 description 17
- 230000033001 locomotion Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 230000032683 aging Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 238000005452 bending Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 238000009434 installation Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000428 dust Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 239000003595 mist Substances 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 238000011897 real-time detection Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 206010008132 Cerebral thrombosis Diseases 0.000 description 1
- 208000028698 Cognitive impairment Diseases 0.000 description 1
- 208000034656 Contusions Diseases 0.000 description 1
- 201000001429 Intracranial Thrombosis Diseases 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 208000010877 cognitive disease Diseases 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000009514 concussion Effects 0.000 description 1
- 230000009519 contusion Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005562 fading Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 210000004872 soft tissue Anatomy 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1116—Determining posture transitions
- A61B5/1117—Fall detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Heart & Thoracic Surgery (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Physiology (AREA)
- Electromagnetism (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Dentistry (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a human body falling detection device and a detection method of a double millimeter wave radar, wherein the device comprises two vertically arranged multichannel millimeter wave radars, a signal processing module and an automatic feature extraction and identification module, wherein the millimeter wave radars are used for detecting the posture information of a human body, the signal processing module is used for processing the posture information of the human body to obtain a fused time-frequency domain image, and the automatic feature extraction and identification module is used for extracting the feature value of the fused time-frequency domain image so as to identify the posture of the human body to obtain the falling information of the human body; the invention has good privacy protection and more accurate detection result for human body falling.
Description
Technical Field
The invention belongs to the technical field of signal and information processing, and relates to a human body fall detection device and a detection method of a double millimeter wave radar.
Background
The current world population aging problem is increasingly aggravated, according to the recently issued aging problem report of the world health organization, the population over 60 years old in 2050 is expected to reach 20% of the world population, particularly Asian countries such as China, the proportion of the aging population is expected to be up to 40%, the population aging brings a prominent social problem, namely the health care problem of the domestic old, according to the statistics of the world health organization, the falling probability of the old over 65 years old exceeds 28% each year, and the falling is the main cause of death of the old over 79 years old, on one hand, the injury caused by the falling, such as fracture, soft tissue contusion and the like, and on the other hand, the complications, such as heart concussion, cerebral thrombosis and the like, caused by the fact that the old cannot be found and seek medical time in time after the old falls are expected, so that an effective method for researching the falling detection of a human body is urgently needed.
The research aiming at human body fall detection is mainly divided into three types, the first type is an indoor fall detection technology based on a wearable sensor, the technology detects human body gestures through various sensors worn on a human body or embedded in the human body so as to judge whether a fall event occurs, and the method has the characteristics of high sensitivity, good real-time detection and the like, however, the method is invasive, and the portable sensor is generally unfriendly to users and is easy to damage, and the button type alarm equipment is not suitable for patients with cognitive impairment or old people; the second type is an indoor falling detection technology based on a video/image sensor, the detection system records a series of human body moving images/videos through a high-resolution camera, and the images/videos are processed by adopting an action classification algorithm, so that falling of a human body can be effectively detected, but the installation of a camera can involve invading privacy, especially areas such as a bathroom, a bathroom and a bedroom, which are places which are extremely easy to fall, and the detection effect of the camera can be influenced by water mist, dust and the like in the environment, so that the application of the detection system is limited to a certain extent; the third type is a fall detection technology based on radar detection of human body movement, which uses the principle that the human body movement can cause the radar echo signal frequency to change (namely Doppler effect) to detect, has the advantages of non-contact, good privacy protection, capability of penetrating fog and the like, and is a fall detection means with very high application potential. However, when the radar echo signal of the falling action of the human body is subjected to time-frequency domain representation, the high-frequency component strength is weaker, and the time-width measurement accuracy of the representation method using short-time Fourier transform under the high Doppler frequency is influenced by the time resolution; in addition, the traditional single millimeter wave radar detection mode has a detection blind area, and when a human body falls along the tangential direction of the radar beam direction, the Doppler domain signal energy is weaker, is seriously affected by noise, and is difficult to identify.
Disclosure of Invention
In order to achieve the above-mentioned purpose, the embodiment of the invention provides a human body falling detection device of double millimeter wave radar, the device has simple structure, can accurately detect the human body falling condition without contacting with human body, has good privacy protection, and can accurately detect the human body falling in complex environments such as water mist, dust and the like.
The embodiment of the invention also provides a human body falling detection method of the double millimeter wave radar, which utilizes a time-frequency domain analysis method to realize radar signal characteristic enhancement for effectively reflecting human body falling actions under the condition of low signal-to-noise ratio, and utilizes time-frequency information fusion of the double millimeter wave radar and an automatic characteristic extraction method based on deep learning to realize accurate identification of human body gestures.
The technical scheme adopted by the embodiment of the invention is that the human body falling detection device with double millimeter wave radars comprises two millimeter wave radars, a signal processing module and an automatic feature extraction and identification module;
the millimeter wave radar is used for transmitting detection signals, receiving echo signals reflected by a human body and preprocessing the echo signals to obtain multipath echo digital signals;
The signal processing module is used for carrying out digital wave beam formation, FFT operation, stockwell transformation and weighted data fusion on the multipath echo digital signals to obtain a fused time-frequency domain image;
The automatic feature extraction and identification module is used for extracting feature values of the fused time-frequency domain image, and further identifying the human body posture to obtain the falling situation of the human body.
Further, the millimeter wave radar is a multi-input multi-output multi-channel millimeter wave radar, and the two millimeter wave radars are placed at 90 degrees in the detection area.
Further, the automatic feature extraction and identification module comprises an image input layer, a middle layer, a full connection layer and an output layer which are sequentially connected;
the image input layer is used for preprocessing the fused time-frequency domain image;
The intermediate layer is used for extracting the characteristics of the preprocessed fused time-frequency domain image;
the full-connection layer is used for classifying and identifying the human body gesture according to the extracted characteristic value;
the output layer is used for outputting the human body gesture classification and identification result.
Further, the middle layer is formed by sequentially connecting a first convolution block, a maximum pooling layer, a second convolution block, a maximum pooling layer and a third convolution block, the first convolution block, the second convolution block and the third convolution block are formed by sequentially connecting a two-dimensional convolution layer, a batch normalization layer and a correction linear unit layer, and the number of filters of the two-dimensional convolution layer is 8, 16 and 32 in sequence.
The human body falling detection method of the double millimeter wave radar comprises the following steps:
step S1, two millimeter wave radars simultaneously transmit detection signals to a detection area, the signals are received by receiving antennas of the millimeter wave radars after being reflected by a human body, the front end of the millimeter wave radar receiver preprocesses echo signals to obtain multipath echo digital signals of the radars, and the multipath echo digital signals are sent to a signal processing module;
Step S2, respectively carrying out digital beam forming on multipath echo digital signals sent by two millimeter wave radars, and sequentially carrying out fast time dimension FFT operation and slow time dimension Stockwell transformation on the digital beams to obtain two human body gesture Doppler domain time-frequency images in the vertical direction;
step S3, carrying out weighted data fusion on the two human body gesture Doppler domain time-frequency images in the vertical direction to obtain a fused time-frequency domain image;
and S4, extracting characteristic values of the fused time-frequency domain image, and classifying and identifying the human body posture based on the characteristic values to obtain the human body falling situation.
Further, the pretreatment process in step S1 is as follows: firstly, the echo signal is amplified with low noise, then mixed with the local oscillation signal, converted to intermediate frequency by radio frequency, and finally, the intermediate frequency sampling is carried out by using a digital-to-analog converter to obtain the multipath echo digital signal of the millimeter wave radar.
Further, in the step S3, the following procedure is used to perform weighted data fusion: and respectively extracting the ridge lines of the two human body gesture Doppler time-frequency domain images, and adding the frequency values of the two ridge lines at the same time to obtain a fused time-frequency domain image.
The embodiment of the invention has the beneficial effects that: 1. according to the embodiment of the invention, two multi-channel millimeter wave radars are used for vertical sampling, so that the problem that a single radar has a detection blind area is avoided, the characteristic data of human body fall detection is increased, and the detection result is more accurate; 2. according to the embodiment of the invention, fast time dimension FFT operation and slow time dimension Stockwell transformation are sequentially carried out on the digital wave beams of the radar echo signals, the defect that the time width of a short-time Fourier transformation window is unchanged is overcome, the analysis time width can be adaptively adjusted according to frequency transformation, visual time-frequency characteristics are provided, the window function domain analysis scale is not required to be selected, the method is more suitable for detecting weak signals (especially high-frequency weak components), the phase information faded in wavelet transformation and short-time Fourier transformation is reserved, the time-frequency and characteristic information extraction of the signals is obviously advantageous, and the identification rate of human body falling can be obviously improved; 3. according to the embodiment of the invention, the Doppler domain information of the two multi-channel millimeter wave radars is weighted and fused by adopting the image ridge line extraction, so that the information easy to identify in the human body falling action is amplified, the distinguishable characteristics in the echo signals are enhanced, the distinguishable characteristics are input into the classifier model based on the convolutional neural network, the human body falling action can be accurately identified, and the identification rate is superior to that of the existing method for predefining and manually designing and extracting the characteristics.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic structural view of the present invention.
Fig. 2 is an application effect diagram of the present invention.
Fig. 3 is a data processing flow chart of the present invention.
Fig. 4 is a CNN model training flow chart of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Examples
As shown in fig. 1, the human body falling detection device of the double millimeter wave radar comprises two millimeter wave radars, a signal processing module and an automatic feature extraction and identification module, wherein the two millimeter wave radars comprise two vertically placed multichannel millimeter wave radars, a computer is used for controlling the two multichannel millimeter wave radars to simultaneously emit electromagnetic waves to a monitoring area and simultaneously receive echo signals emitted by a human body, and the echo signals are preprocessed and then transmitted to the signal processing module.
The signal processing module performs FFT operation and Stockwell transformation on the received echo signals, enhances the characteristics after time-frequency analysis, fuses multi-channel multi-azimuth information of two multi-channel millimeter wave radars, and transmits the fused information to the automatic characteristic extraction and identification module; the automatic feature extraction and recognition module firstly extracts the feature value of the fusion image, and then classifies and recognizes the signals by using a classification algorithm based on a convolutional neural network, so as to judge the motion state of the human body.
The multi-channel millimeter wave radar adopts a multi-input multi-output (MIMO) working mode such as 2-transmission 4-reception and the like, is provided with a plurality of transmitting antennas and receiving antennas, and the virtual aperture equivalent to the multiple transmission and the multiple reception can greatly improve the effective aperture of the array, thereby effectively improving the spatial resolution of the radar and accurately identifying the motion state of a human body; as shown in fig. 2, two multi-channel millimeter wave radars are placed at 90 degrees in a detection area, the installation height is 1m, the installation is parallel to a wall surface, the horizontal emission angle of an antenna is 0 degree, electromagnetic waves are ensured to be vertically emitted to the detection area, the problem that a single radar has a detection blind area can be better solved, and the characteristic extraction performance is improved.
The parameters of the two multi-channel millimeter wave radars are set consistently, connection is established between the two multi-channel millimeter wave radars when computer control is used, and the communication between the two multi-channel millimeter wave radars is realized by adopting hard triggering, so that the synchronism of data acquisition and the analyzability of frequency domain information fusion are ensured, and the accuracy of detection and identification results of human body falling is good.
As shown in fig. 3, the human body fall detection method of the double millimeter wave radar comprises the following steps:
step S1, two multi-channel millimeter wave radars transmit signals to a detection area at the same time, the signals are reflected by a human body and then are transmitted back to a receiving antenna of the multi-channel millimeter wave radars, and echo signals are subjected to down-conversion processing and then are converted by a digital-to-analog converter (ADC) to obtain multi-channel echo digital signals;
the specific process is as follows: firstly, an antenna receives an echo signal, then the echo signal passes through a low noise amplifier, then the echo signal is mixed with a local oscillator signal through a mixer, then the radio frequency is converted into an intermediate frequency, then an ADC (analog to digital converter) performs intermediate frequency sampling, and a plurality of paths of echo digital signals of a radar are obtained after sampling; the process is completed by the front end of the radar receiver;
step S2, after receiving multipath echo digital signals of two multipath millimeter wave radars, a signal processing module respectively processes the multipath echo digital signals;
firstly, digital beam forming is carried out on multi-path echo digital signals, data are converted from an array element domain to a beam domain, the detection range of a radar can be enlarged by utilizing a digital beam forming technology, and non-blind area real-time detection is carried out on a detection area better;
Performing fast-time-dimension FFT operation on the echo signals, performing slow-time-dimension Stockwell transformation on the processed signals, and obtaining Doppler domain time-frequency information of human body motion from two vertical directions; firstly, carrying out FFT operation on a single period Chirp echo sequence, extracting the abscissa frequency corresponding to a frequency spectrum peak value from each frame of data in a fast time dimension, completing the solution of a target distance, and then carrying out Stockwell conversion on the Chirp echo sequence of a same distance unit after accumulating a plurality of Chirp echo sequences to obtain two-dimensional data information with the frequency changing along with time; the Stockwell transform formula is as follows:
Wherein S represents Stockwell transformation, t represents time, x (t) represents analysis signals at the moment t, tau represents the position of a window function on a time axis, f represents frequency, S (tau, f) represents a time spectrum matrix obtained by transformation, and i represents a complex imaginary part;
The slow time transformation is carried out by adopting a short time Fourier transformation mode, the frequency resolution is improved and the time resolution is reduced when the selected window length is prolonged, the time resolution is improved and the frequency resolution is reduced when the selected window length is shortened, the Stockwell transformation is based on an expandable local Gaussian window and provides frequency-dependent resolution, the slow time transformation is the variable window transformation of the Short Time Fourier Transformation (STFT) and the expansion of the wavelet transformation, and is suitable for the detection of weak signals, particularly has the enhancement effect on high-frequency weak components, and meanwhile, the phase information attenuated in the wavelet transformation and the short time Fourier transformation is reserved, so that the slow time Fourier transformation is more suitable for the falling detection of human bodies;
S3, carrying out weighted data fusion on Doppler domain time-frequency information in two directions by adopting a method for extracting data fusion by using an image ridge line to obtain fused time-frequency domain information;
the data fusion process is as follows: respectively extracting the ridge lines of the two time-frequency distribution graphs obtained by echo processing of the two millimeter wave radars to obtain a time-frequency change track with characteristic information contained in the time-frequency domain information, wherein the ridge lines extracted by the two graphs are corresponding in time because the data acquisition processes of the two millimeter wave radars are triggered simultaneously;
The time axis is taken as an X axis, and the frequency axis is taken as a Y axis, namely, the frequency values of two ridge lines at the same moment are added, so that information easy to identify in the falling action of a human body is amplified, the distinguishable characteristics in echo signals are effectively enhanced, the effective frequency transformation becomes more remarkable, and the subsequent characteristic extraction and identification are facilitated;
And S4, extracting a characteristic value of the fused time-frequency domain information by the automatic characteristic extraction and identification module, classifying and identifying the human body posture based on the characteristic value, and judging the motion state of the human body.
As shown in fig. 4, the automatic feature extraction and recognition module is composed of an image input layer, an intermediate layer, a full-connection layer and an output layer which are sequentially connected, the image input layer pre-processes the input fused time-frequency domain information and then inputs the fused time-frequency domain information into the intermediate layer, the intermediate layer performs multi-level feature extraction on the input time-frequency distribution map, then the full-connection layer is used for re-fitting the extracted features, loss of feature information is reduced, the human body posture is classified and recognized based on the human body posture features, and the human body falling condition is output through the output layer.
The image input layer normalizes the data of the fused time-frequency domain information (namely the synthesized two-dimensional time-frequency distribution diagram), then inputs the fused time-frequency domain information into the middle layer, and the human body movement gesture comprises a falling action and a non-falling action, wherein the falling action comprises forward falling, backward falling and forward and backward falling and is added with arm support, and the non-falling action comprises bending, rapid bending and squatting;
The middle layer is formed by sequentially connecting a first convolution block, a maximum pooling layer, a second convolution block, a maximum pooling layer and a third convolution block, wherein the first convolution block, the second convolution block and the third convolution block comprise sequentially connected two-dimensional convolution layers, batch normalization layers and correction linear unit layers, and the number of filters of the three two-dimensional convolution layers is 8, 16 and 32 in sequence; the two-dimensional convolution layer of the first convolution block calculates dot products of weight and input by moving a filter vertically and horizontally along an input direction, then adds offset items to convolve the input, then enters a batch normalization layer to normalize, then inputs a correction linear unit layer to execute threshold operation on each element, finally inputs a maximum pooling layer, the maximum pooling layer divides the input into rectangular pooling areas, calculates the maximum value of each pooling area to perform downsampling, then sequentially processes the maximum value through a second convolution block, the maximum pooling layer and a third convolution block, inputs a processing result into a full connection layer, multiplies the input by a weight matrix by the full connection layer, then adds offset vectors to obtain high-dimensional space data, reduces loss of characteristic information, and the output layer is used for outputting a human motion state recognition result.
When training the automatic feature extraction and recognition module, firstly, collecting a large number of echo data sets of human body gestures under the situations of falling and non-falling, carrying out data processing and weighted fusion on the echo data sets, then storing fused time-frequency domain images, using the fused time-frequency domain images as training samples, training the automatic feature extraction and recognition module, and verifying the automatic feature extraction and recognition module by using the echo data of the human body gestures under the situations of new falling and non-falling to obtain the final automatic feature extraction and recognition module.
The method comprises the steps of respectively collecting action data of 60 volunteers as a data set, collecting 14 kinds of actions in total by each volunteer, wherein the first group of actions are forward falling, backward falling, forward falling and arm supporting, backward falling and arm supporting, bending down, squatting and sitting down towards a radar, the second group of actions are 7 actions towards the radar at an angle of 45 degrees, and the human body falling detection is carried out on the collected data by using the method and the traditional detection method, and the detection results are shown in the table 1:
Table 1 identification rate data of the present invention and conventional detection method for human fall detection
As can be seen from table 1: according to the traditional method for analyzing the time-frequency of the short-time Fourier transform of the single millimeter wave radar, the recognition rate of the single millimeter wave radar data on the falling action can reach 88% by combining with the trained convolutional neural network, if the single millimeter wave radar data is analyzed by using the Stockwell transform, the recognition rate can be improved to 93%, the recognition accuracy of the falling detection is greatly improved by the method of the Stockwell transform, the two millimeter wave radar data which are vertically placed are subjected to FFT operation and Stockwell transform and then fused with time-frequency domain information and then are subjected to classification recognition, the recognition rate can reach 96%, the action data of a human body can be detected without dead zone by using the double millimeter wave radar, the time width of the analysis can be adjusted according to frequency adaptation by the FFT operation and Stockwell transform, the high-frequency weak component in an echo signal is enhanced, the phase information of fading in the wavelet transform and the short-time Fourier transform is reserved, the time frequency and characteristic information of the echo signal are conveniently extracted, the human body falling detection is carried out on the basis of the extracted characteristics, and the detection result recognition rate is higher.
In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
The foregoing description is only of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.
Claims (5)
1. The human body falling detection device of the double millimeter wave radar is characterized by comprising two millimeter wave radars, a signal processing module and an automatic feature extraction and identification module;
the millimeter wave radar is used for transmitting detection signals, receiving echo signals reflected by a human body and preprocessing the echo signals to obtain multipath echo digital signals;
The signal processing module is used for carrying out digital wave beam formation, FFT operation, stockwell transformation and weighted data fusion on the multipath echo digital signals to obtain a fused time-frequency domain image;
the automatic feature extraction and identification module is used for extracting the feature value of the fused time-frequency domain image, and further identifying the human body posture to obtain the falling condition of the human body;
the automatic feature extraction and identification module comprises an image input layer, an intermediate layer, a full-connection layer and an output layer which are sequentially connected;
the image input layer is used for preprocessing the fused time-frequency domain image;
The intermediate layer is used for extracting the characteristics of the preprocessed fused time-frequency domain image;
the full-connection layer is used for classifying and identifying the human body gesture according to the extracted characteristic value;
The output layer is used for outputting the human body gesture classification and identification result;
The middle layer is formed by sequentially connecting a first convolution block, a maximum pooling layer, a second convolution block, a maximum pooling layer and a third convolution block, the first convolution block, the second convolution block and the third convolution block are formed by sequentially connecting a two-dimensional convolution layer, a batch normalization layer and a correction linear unit layer, and the number of filters of the two-dimensional convolution layer is 8, 16 and 32 in sequence.
2. The dual millimeter wave radar human body fall detection device according to claim 1, wherein the millimeter wave radar is a multiple-input multiple-output multichannel millimeter wave radar, and the two millimeter wave radars are disposed at 90 ° in the detection area.
3. A method of detecting a fall of a person using a device as claimed in claim 1 or 2, comprising the steps of:
step S1, two millimeter wave radars simultaneously transmit detection signals to a detection area, the signals are received by receiving antennas of the millimeter wave radars after being reflected by a human body, the front end of the millimeter wave radar receiver preprocesses echo signals to obtain multipath echo digital signals of the radars, and the multipath echo digital signals are sent to a signal processing module;
Step S2, respectively carrying out digital beam forming on multipath echo digital signals sent by two millimeter wave radars, and sequentially carrying out fast time dimension FFT operation and slow time dimension Stockwell transformation on the digital beams to obtain two human body gesture Doppler domain time-frequency images in the vertical direction;
step S3, carrying out weighted data fusion on the two human body gesture Doppler domain time-frequency images in the vertical direction to obtain a fused time-frequency domain image;
and S4, extracting characteristic values of the fused time-frequency domain image, and classifying and identifying the human body posture based on the characteristic values to obtain the human body falling situation.
4. A method according to claim 3, wherein the pretreatment in step S1 is as follows: firstly, the echo signal is amplified with low noise, then mixed with the local oscillation signal, converted to intermediate frequency by radio frequency, and finally, the intermediate frequency sampling is carried out by using a digital-to-analog converter to obtain the multipath echo digital signal of the millimeter wave radar.
5. A method according to claim 3, characterized in that in step S3 the weighted data fusion is performed using the following procedure: and respectively extracting the ridge lines of the two human body gesture Doppler time-frequency domain images, and adding the frequency values of the two ridge lines at the same time to obtain a fused time-frequency domain image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110728497.9A CN113447905B (en) | 2021-06-29 | 2021-06-29 | Double millimeter wave radar human body fall detection device and detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110728497.9A CN113447905B (en) | 2021-06-29 | 2021-06-29 | Double millimeter wave radar human body fall detection device and detection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113447905A CN113447905A (en) | 2021-09-28 |
CN113447905B true CN113447905B (en) | 2024-04-19 |
Family
ID=77814007
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110728497.9A Active CN113447905B (en) | 2021-06-29 | 2021-06-29 | Double millimeter wave radar human body fall detection device and detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113447905B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113793478A (en) * | 2021-10-11 | 2021-12-14 | 厦门狄耐克物联智慧科技有限公司 | Microwave induction toilet tumble alarm system |
CN114366052A (en) * | 2021-12-21 | 2022-04-19 | 山东师范大学 | Intelligent nursing home monitoring system and method based on millimeter wave radar |
CN114584747B (en) * | 2022-03-04 | 2023-10-31 | 大连海事大学 | 360-degree annular curtain seamless projection soft correction method |
CN114924249B (en) * | 2022-07-22 | 2022-10-28 | 中国科学技术大学 | Millimeter wave radar-based human body posture estimation method and device and electronic equipment |
WO2024044870A1 (en) * | 2022-08-29 | 2024-03-07 | 上海金桥(集团)有限公司 | Underground parking lot passage moving vehicle real-time detection system and method |
CN115291184B (en) * | 2022-10-08 | 2023-03-24 | 四川启睿克科技有限公司 | Attitude monitoring method combining millimeter wave radar and deep learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111166342A (en) * | 2020-01-07 | 2020-05-19 | 四川宇然智荟科技有限公司 | Millimeter wave radar and camera fused fall detection device and detection method thereof |
CN112617813A (en) * | 2020-12-15 | 2021-04-09 | 南京邮电大学 | Multi-sensor-based non-invasive fall detection method and system |
CN112782681A (en) * | 2020-12-31 | 2021-05-11 | 杭州电子科技大学 | Indoor positioning and falling detection system and method based on millimeter waves and Internet of things |
WO2021109526A1 (en) * | 2019-12-03 | 2021-06-10 | 南京汇君半导体科技有限公司 | Vital sign signal enhancement method and device, and extraction method and device based on millimeter-wave radar |
-
2021
- 2021-06-29 CN CN202110728497.9A patent/CN113447905B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021109526A1 (en) * | 2019-12-03 | 2021-06-10 | 南京汇君半导体科技有限公司 | Vital sign signal enhancement method and device, and extraction method and device based on millimeter-wave radar |
CN111166342A (en) * | 2020-01-07 | 2020-05-19 | 四川宇然智荟科技有限公司 | Millimeter wave radar and camera fused fall detection device and detection method thereof |
CN112617813A (en) * | 2020-12-15 | 2021-04-09 | 南京邮电大学 | Multi-sensor-based non-invasive fall detection method and system |
CN112782681A (en) * | 2020-12-31 | 2021-05-11 | 杭州电子科技大学 | Indoor positioning and falling detection system and method based on millimeter waves and Internet of things |
Non-Patent Citations (1)
Title |
---|
视觉和毫米波雷达信息融合行人识别算法;徐伟;周培义;张芬;黄亮;;同济大学学报(自然科学版);20171215(第S1期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN113447905A (en) | 2021-09-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113447905B (en) | Double millimeter wave radar human body fall detection device and detection method | |
CN110286368B (en) | Old people falling detection method based on ultra wide band radar | |
CN109507653A (en) | A method of multi-information perception bioradar system and its acquisition target information based on UWB | |
CN109672485B (en) | Indoor personnel real-time invasion and movement speed detection method based on channel state information | |
CN108038419B (en) | Wi-Fi-based indoor personnel passive detection method | |
CN112686094B (en) | Non-contact identity recognition method and system based on millimeter wave radar | |
CN112333653A (en) | Identity intelligent identification method and system based on WiFi channel state information | |
CN112861813B (en) | Method for identifying human behavior behind wall based on complex value convolution neural network | |
CN107462873A (en) | A kind of radar chaff method for quickly identifying | |
Wang et al. | Multi-classification of UWB signal propagation channels based on one-dimensional wavelet packet analysis and CNN | |
CN113453180B (en) | Intelligent detection method and system for human body tumble and information data processing terminal | |
CN114814775A (en) | Radar fall detection method and equipment based on ResNet network | |
CN109431526B (en) | WIFI-based driving state identification method and system | |
CN114814832A (en) | Millimeter wave radar-based real-time monitoring system and method for human body falling behavior | |
CN111901028B (en) | Human body behavior identification method based on CSI (channel State information) on multiple antennas | |
WO2023284764A1 (en) | Method and apparatus for detecting living body in vehicle by means of radar, and terminal device | |
CN112147584A (en) | MIMO radar extended target detection method based on non-uniform clutter | |
Bocus et al. | UWB and WiFi systems as passive opportunistic activity sensing radars | |
Showmik et al. | Human activity recognition from wi-fi csi data using principal component-based wavelet cnn | |
Qian et al. | Parallel LSTM-CNN network with radar multispectrogram for human activity recognition | |
CN113341392B (en) | Human behavior classification method based on multi-station radar micro-Doppler motion direction finding | |
CN117315886B (en) | UWB radar-based method and device for detecting impending falling of personnel | |
KR102322480B1 (en) | Position Detecting System and Method of Ultra Wide Band Using the Optimized Detection Threshold at Objects Size and Motion Strenth | |
CN115575921B (en) | Pitching-direction-based multichannel multi-interference-base suppression interference suppression method | |
CN111329447B (en) | Human body physiological parameter detection method and device and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |